FreeNet, *ster, and horizontal rumblings...

Date view Thread view Subject view Author view

From: Jeff Bone (
Date: Wed Mar 22 2000 - 21:59:59 PST

...or make that "rumblings on the horizon." My ENTPish seis-meme-meter (TM!)
is shaking somewhat, feels kind of like those quiet, mid-96 distributed caching
tremors but somehow different.

I just read Ian Clarke's (the FreeNet guy's) academic paper [1] on adaptive
networking and I'm kind of baffled. Anybody else know anything about FreeNet?
On the one hand, it's clear that the *ster meme has truly set in. And there's
some interesting corollary memes that FreeNet captures: anonymity, etc. But
to me, this particular effort looks a lot like a rehash of the whole "how can
we efficiently implement distributed Linda?" thing circa mid-80s.

OTOH, I do think there're some powerful ideas running around this whole *ster
space. In its purest form, it is pure, decentralized, opportunistic, indexed,
anonymized object (file) sharing / caching. It's Battle of the Big Memes, sort
of an Open Source redux: while Open Source is "cathedral vs. bazaar" and to a
lesser sense a pure anti-IP thing (and that only in the fringes,) it looks to
me like *ster is really _all_ about "infoconsumers vs. infoproducers," or more
specifically "economics of abundance" vs. "economics of (artificial) scarcity."

It will be interesting to watch all of this take shape. "Politics" shaping
architecture. How cool is that? :-)

Rough cut segueway...

Interesting intersections of recent and long-term obsessions: Lifestreams;
anonymous storage; the total absolute suckage of bookmarks / histories / etc;
total location transparency; distributed FSes that support disconnected
operation; the lack of decent high-level abstractions (i.e., types, triggers,
queries on just about anything other than naming closures, etc.) in
filesystems; attributed filesystems ala Nebula and Synopsis; category theory
and concept theory; separation of directories and storage; directories as
continuous queries; gossip and other agoric information-sharing techniques;
opportunistic filesystems; heavy indexing; immutable files as the basis of
distributed filesystems (ala Amoeba's Bullet Server approach.)

The architectural memes and the frustrated user memes are starting to intersect
in interesting ways for me lately. The "product," if it ever actually happens,
is something we referred to for years at Active* as "Data Heaven." [2] I'm
pretty sure we've all been dreaming essentially the same dream for a long
time. (I distinctly recall strong resonances when I looked at Dan Connolly's
Web research notebook. Hi Dan! [3]) If this tickles anyone else's fancy, drop
me a line. Modulo bandwidth and constraints -wrt- the newco (which is *not*
addressing this problem) I've been trying to figure out how to get momentum
going on this front, even to the point of trying to figure out how to "seed" an
open source effort towards this. (So many things I want to exist, so little
time... ;-)



PS: clearly, this doesn't require a new protocol. HTTP is a great transport.
PPS: check this guy [4] out. Anybody know him? Looks like Martin Short's
geeky brother, if that groks properly. :-) Good pointers, though.

[2] You know the story: all my data, from anywhere, connected and
disconnected operation, synchronization, nothing deleted, auto-archiving, heavy
indexing, the machine does the heavy lifting with filing / finding, all data
created / treated equally, supports sharing / collaboration, good framework for
agency / knowledge engineering, etc. etc.

Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed Mar 22 2000 - 22:05:21 PST